Computer Vision Developer for Photo Camera Localization, Change Detection

We are seeking an experienced Computer Vision developer to complete a challenging project involving old photo camera localization, change detection, and the implementation of learned geometry methods. We currently have existing code, but it requires debugging and further development to function correctly, especially concerning COLMAP integration.

The project deliverables and evaluation criteria are as follows:

1. Slides and Oral Talk (10% of overall project):
    * Prepare and upload slides for a technical presentation.
    * Deliver an oral talk, including reserving the necessary slot.

2. Quantitative Information (10% of overall project):
    * Provide comprehensive tables reporting quantitative metrics such as RMSE errors, computing time, or number of matches for all implemented methods.

3. Old Photo Camera Localization (40% of overall project):
    * Proposed Method (10%): Develop and implement a robust method for localizing old photo cameras.
    * Experimental Performance (10%): Evaluate the experimental performance of the proposed localization method.
    * Comparison with COLMAP (10%): Conduct a detailed comparison of the proposed method against COLMAP.
    * COLMAP with New Images (10%): Implement and evaluate COLMAP using 10 or more new images for comparison.

4. Determining Changes (20% of overall project):
    * Proposed Method (10%): Develop and implement a method for detecting changes between images.
    * Experimental Performance (10%): Evaluate the experimental performance of the proposed change detection method.

5. Usage of Learned Geometry Methods (20% of overall project):
    * Integrate and utilize a learned geometry method such as DUSt3R, MASt3R, or VGGT.
    * Results of the Method (10%): Present the results obtained from the chosen learned geometry method.
    * Quantitative Comparison (10%): Provide a quantitative comparison of the learned geometry method with respect to your solution and/or COLMAP.

The ideal candidate will have strong expertise in computer vision, particularly with 3D reconstruction, localization, and deep learning-based geometry methods. Experience with debugging and optimizing existing codebases is essential. We are looking for a freelancer who can not only deliver functional code but also provide thorough analysis and clear presentations of the results.